Senior Backend Engineer — Scheduler
Lead scheduling-policy work on the CNLab federated orchestrator. Strong distributed systems background.
Team: Platform Engineering· Manager: CTO· Comp: Negotiable
About the role
CNLab's scheduler is the brain of the platform. It decides where every workload runs — across MIG slices, 1% blocks, multiple clusters, and (sometimes) cloud burst. We're hiring a senior engineer to lead the next generation of scheduling policies, including better deadline handling, smarter pre-emption, and tighter integration with our predictive burst-out forecaster.
What you'll do
- Design and implement new scheduling policies (priority + deadline + fair-share + bin-packing combinations).
- Own the placement decision pipeline end-to-end: queue management, candidate scoring, placement, and rollback on infeasibility.
- Work with the research team to integrate the predictive burst-out forecaster into the live scheduler.
- Define SLOs for scheduling decisions (median, P95, P99 placement latency under load) and own the runbook for breach.
- Mentor two mid-level engineers on the same team.
What we're looking for
- 5+ years of backend engineering, with 2+ on a real scheduler (Kubernetes scheduler, Slurm, YARN, custom).
- Strong distributed systems fundamentals — consensus, time, fault models, capacity planning.
- Production experience with Go or TypeScript (we use both; this role leans TypeScript via NestJS).
- Comfortable working with researchers — translate ideas into code that handles edge cases.
- Excellent written communication. We write a lot of internal docs.
Bonus
- Experience with GPU schedulers specifically.
- Open-source contributions to Kubernetes, Volcano, Yunikorn, or Kueue.
- Korean language (not required — half the team works in English daily).
Comp & benefits
Salary is highly negotiable and aligned with industry standards and expertise.
Process
- Apply — short application form. Resume optional; describe a hard scheduling problem you've shipped.
- 30-min intro chat with the hiring manager.
- System-design interview — 90 min, real-world scheduler design.
- Code-review pairing — 60 min, you walk us through a recent piece of your code.
- On-site (or remote-equivalent): 4 conversations, including a 60-min architecture deep-dive with the CTO.
- Offer.
End-to-end timeline: 10–15 business days. We make decisions quickly.